Concerted elimination of Cl + 2 from CCl 4 and of I + 2 from CH 2I 2 driven by intense ultrafast laser pulses

Transcription

1 Concerted elimination of Cl + 2 from CCl 4 and of I + 2 from CH 2I 2 driven by intense ultrafast laser pulses A Thesis Presented by Dominik Peter Michael Geissler to The Graduate School in Partial Fulfillment of the Requirements for the Degree of Master of Arts in Physics Stony Brook University December 2007

2 Stony Brook University The Graduate School Dominik Peter Michael Geissler We, the Thesis committee for the above candidate for the Master of Arts degree, hereby recommend acceptance of the Thesis. Thomas C. Weinacht, Thesis Advisor Assistant Professor, Department of Physics and Astronomy Dominik A. Schneble Assistant Professor, Department of Physics and Astronomy Michael A. Zingale Assistant Professor, Department of Physics and Astronomy This Thesis is accepted by the Graduate School. Lawrence Martin Dean of the Graduate School ii

3 Abstract of the Thesis Concerted elimination of Cl + 2 from CCl 4 and of I + 2 from CH 2I 2 driven by intense ultrafast laser pulses by Dominik Peter Michael Geissler Master of Arts in Physics Stony Brook University 2007 The concerted elimination of Cl + 2 from CCl 4 and I + 2 from CH 2 I 2 is studied with various methods. Single pulse experiments proved the possibility of Cl + 2 elimination from CCl 4 but were limited by background contamination. Pulse shape varying experiments (with use of a genetic algorithm) demonstrated a pulse shape dependent concerted elimination of I + 2 from CH 2 I 2. Furthermore it was possible to partly resolve the involved molecular states for this elimination process by pump probe scans both in the red as in the UV spectrum. We could follow the evolution of a vibrational wave packet in the halogenated methane iii

4 CH 2 I 2 and observed how the wave packet modulates both dissociation and concerted elimination to form CH 2 I + and I + 2 respectively. Finally a simple and intuitive interpretation of the molecular dynamics leading to the formation of the products is presented. iv

5 To Martin and my family

6 Contents List of Figures... x Acknowledgements... xi 1 Introduction The experimental apparatus The laser system The interaction chamber The red-red pump-probe alignment The detector and data acquisition Frequency Resolved Optical Gating The UV-Red setup Ultrafast Optics Dispersion Control vi

7 3.2 Nonlinear optical effects Ultrafast coherent control and closed loop feedback with genetic algorithm CCl 4 experiments General facts of the Cl 2 data Genetic Algorithm Experiments Diiodomethane CH 2 I 2 experiments Genetic Algorithm experiments Red-Red pump probe experiments UV-Red pump probe experiments Discussion of the CH 2 I 2 pump-probe data The oscillations in the ion yield The left hand side of the red-red data Frequency and phase of the oscillations Potential energy surfaces Candidates for the intermediate state CH 2 I r Neutral states An excited ionic state Limits of the one-dimensional picture vii

8 6.6 UV-Red pump-probe scan Conclusions Bibliography viii

9 List of Figures 2.1 Schematic view of the pulse shaper Schematic view of the red experimental setup Schematic view of vacuum system Example trace of the TOFMS taken by the oscilloscope UV-Red pump-probe setup Red to UV tripling setup Schematic diagram of a prism compressor Sum frequency generation Closed loop learning control Fitness as a function of generation Schematic view of a CCl 4 molecule. It has a methane-like shape, whereby the hydrogen atoms are replaced with chlorine atoms Ion yield of fragmented CCl 4 for pulse energies from 0 to 210µJ TOFMS of CCl 4 at 220µJ pulse energy ix

10 4.4 TOFMS of CCl 4 with an unshaped / optimal shaped pulse Shape of a pulse with maximal Cl + 2 yield Relative GA improvement at different energies Yields of several fragments derived from CCl 4 as a function of pulse energy Structure of the CH 2 I 2 molecule Ion yield of fragmented CH 2 I 2 for pulse energies from 0 to 210µJ Optimal pulse shape for maximum I + 2 with a CH 2 I 2 sample Ion signal of a red-red pump probe scan with CH 2 I Ion signal of a red-red pump probe scan with pure iodine Ion signal of a UV-red pump probe scan with CH 2 I Fourier analysis of the oscillations in different ion yields Normalized signal of several fragments versus delay time Cartoon of the involved potential energy surfaces in CH 2 I Oscillation amplitude of several fragments in dependence on the probe pulse energy Molecular structure of Iso-Diiodomethane Normalized ion yields for several UV pump energies Normalized Fourier transform of the UV pumped CH 2 I + 2 signal 78 x

11 Acknowledgements First, I would like to thank my advisor Tom Weinacht for giving me the opportunity to work with him. He supported my work in many different ways: He introduced me to the lab, there have been lot of fruitful discussions and he spread an enthusiasm that made all the long hours and disappointments look minuscule. I also want to thank Sarah Nichols and Brett Pearson for their sher infinite patience to answer my questions and helping me out with my experiment many times. Also all other members of the group Carlos, Stephen, Coco and Marija have to be mentioned here, as I enjoyed working together with them very much. Furthermore I want to thank everyone with whom I spent great times outside the lab and who made my stay an integrated experience. I owe a debt of gratitude to everyone who was involved in the exchange program between the physics departments of the University of Wrzburg and Stony Brook University for all their effort to organize my stay here. A scholarship

12 from the German Merit Foundation (Studienstiftung des deutschen Volkes) provided the financial framework and is highly appreciated, too. In addition, this work was supported by the National Science Foundation. My warmest thanks go to my family for always encouraging me. Without their support I would not have been able to come here and begin this work, nor finish it. Last but not least, I would like to thank my boyfriend Martin for his love and encouragement.

13 Chapter 1 Introduction The idea of generating short laser pulses to observe short processes in nature came up soon after the invention of the laser. Since then vast progress has been made in shortening the pulse duration and increasing the pulse power. Pulse durations in the sub-femtosecond time scale have been reported [11]. Current (2005) records in energy are high power 10 fs pulses at 10.5 mj [41]. In the meantime ultrafast laser systems are available for a continuous spectrum from near IR to UVC for their central wavelength. They have broad application to observe processes that happen on a femtosecond timescale, like atomic/molecular transitions or reactions and solid-state phenomena. At the same time their high intensity surpasses the characteristic fields in atoms or molecules, i.e. E> e 4πɛ 0 r Bohr. The progress in laser technology has been accompanied by the development of techniques for pulse shaping, allowing control 1

14 over the characteristic variables of the electric field: phase, amplitude, and polarization. This opened the field of ultrafast control to interact with atomic and molecular processes. Progress has been made in controlling selective atomic and molecular excitation and selective molecular fragmentation [4, 9, 14, 40]. Femtosecond lasers are ideal to study vibrational wave packet dynamics. They can be used for molecular imaging and play an important role in driving coherent chemistry [1, 18, 35, 37, 42]. Several diatomic molecules have been studied so far, such that research is proceeding to larger molecules. Halogenated methanes make up a broad field of interesting candidates. With them several experiments ([31], [43] and [22]) have been done in examining the control over concerted elimination and bond breaking. In this thesis some experiments will be described and discussed which examine the concerted elimination of molecular halogens from halogenated methanes under femtosecond laser excitation. The model systems CH 2 I 2 and CCl 4 have been studied, i.e. reactions like: CCl 4 + nhν Cl CH 2 I 2 + nhν I There are principally two ways to emit molecular particles from the parent molecule: (a) both halogen atoms are emitted separately and combine afterwards, if their pathes cross or (b) both halogen atoms are emitted and 2

15 combine at the same time. Latter process is referred to concerted elimination (this definition follows [23]) and is directly dependent on the vibrational and rotational motion of the molecule. The aim of this thesis is to develop a simple picture of the involved molecular transitions for the concerted formation of the halogen molecules and to study the controllability of this process. Therefore different approaches have been used: Single pulse experiments ( genetic algorithm experiments ) in combination with a pulse shaper payed attention to the influence of different pulse shapes, double pulse ( pump-probe ) experiments examined the time dynamics of this process and experiments with mixed UV and IR laser searched for an influence of the laser wavelength. 3

16 Chapter 2 The experimental apparatus 2.1 The laser system In the experiments a high power Ti:Sapphire laser system is used, as it is described in [3]. It consists of a Ti:Saphire oscillator, that emits 30 fs laser pulses with a wavelength of about 780 nm at a repetition rate of 85 khz and an energy of about 6 nj. The oscillator is pumped by a commercial VERDI laser (with 532 nm wavelength). These pulses are used as a seed for the Ti:Sapphire amplifier (pumped by a commercial Quantronix laser at 532 nm). Finally the amplifier generates pulses with a time duration of about 30 fs with a repetition rate of 1 khz. The pulse energy reaches up to 1 mj. The spectrum can be described best as a sum of two gaussian curves, in which the main Gaussian distribution around 780 nm usually has a bandwidth of about 30 nm. 4

17 For this project basically two different setups are used. The first one has been built by several members of the group before and uses one ( GA experiments ) or two red pulses ( red/red pump-probe experiments ) that interact with a cloud of molecules inside a vacuum chamber. This setup is roughly described in the following paragraphs. The second one involves double pulse experiments with a red and a UV pulse ( UV/red pump-probe experiments ) instead. This new setup is described in more detail in section 2.6. For some of the experiments the laser pulse is modified in our pulse shaper (for a schematic view see 2.1). Here the phase (in frequency domain) of the pulses is modified, as well as their overall intensity. At the heart of the pulse shaper are two gratings and the accusto-optical-modulator (AOM crystal). The first grating splits up the beam into its different frequencies, i.e. it maps different frequencies to different propagating angles. Then this bundle is reflected by a curved mirror and focussed inside the AOM crystal, so that now the different angles (and herewith frequencies) are mapped to different points inside the AOM. A piezzo crystal is attached to the side of the AOM. It can be modulated by a radio frequency signal (around a carrier frequency of 150 MHz) generated by one of our lab computers. Thereby it launches a sound wave inside the AOM and thus writes a diffraction pattern in the crystal. The phase of this pattern determines the final phase of the diffracted laser pulse. The 5

18 3600 g/mm Grating 2 Curved Mirror 2 Fold Mirror 2 AOM P 150 MHz, from digital waveform board Curved Mirror 1 Fold Mirror 1 Grating 1 Figure 2.1 Schematic view of the pulse shaper: The piezzo crystal (P) is controlled by the radio signal. This excites sonic waves inside the accusto-opticalmodulator (AOM). amplitude of the sound wave influences the intensity of the diffracted light. This process is triggered by the pump laser and the RF signal and can be considered constant for the time of the laser pulse passing through the crystal. The second grating and curved mirror pair reverses the spectral mapping in space of the first one and recombines all frequencies (now with modified phases) to one beam. The pulse shaper has an overall efficiency of about 30% and can write a maximum phase difference of 60π across the whole bandwidth of the laser. 6

19 For some experiments two laser pulses are used. The setup resembles a Mach-Zehnder-Interferometer. After a beam splitter, these pulses follow different paths. One of them passes through the pulse shaper and thus can be modified. The other is sent to a delay stage. This stage is equipped by a motor, that can be controlled by the lab computer. The lab software has control over the position of the stage and thus the path length for the second beam line. Herewith the time difference in delay time of both pules at the chamber can be varied automatically. The minimal step size of the stage corresponds to a resolution of their time difference of 0.33 fs. Finally both beams are focussed inside the vacuum chamber by a lens. Usually for these experiments both pulses have an energy of about 200 µj. Figure 2.2 summarizes the experimental laser setup described so far. 2.2 The interaction chamber All experiments are performed inside a vacuum chamber. Figure 2.3 shows a schematic view of the vacuum chamber. The interaction chamber itself is pumped by a turbo pump that establishes a vacuum pressure of about 10 6 Torr. The molecules are injected from the sample manifold through a small nozzle. All substances discussed in this thesis have sufficient vapor pressure (at 7

20 KM Labs Ti:Sa Oscillator KM Labs HAP-AMP Quantronix DP 570 (ND-YLF) Verdi V5 TOF Spectrometer Pulse Shaper Figure 2.2 Schematic view of the red experimental setup: The green lines represent the 532 nm pump beam, the red line the used 780 nm beam line, and the dark red line the modified pulses by the amplifier. (figure from [20]) room temperature) to receive a sufficient signal. A small reservoir between the sample manifold and the chamber (with separate valves to open) makes sure that the pressure in the molecular beam is independent of the actual vapor pressure in the sample holder, but only depends on the number of molecules inside the reservoir. Furthermore the reservoir is big enough compared to the injection nozzle to guarantee a molecular beam with constant pressure. Only in long time experiments (> 1 hour) a linear pressure decay becomes noticeable. The final working pressure in the chamber with some sample molecules is usually around 10 5 Torr. The fact that a molecular beam is used has two convenient consequences: (a) Each laser pulse sees a new cloud of sample 8

21 a Vacuum Valve B To Mechanical Pump PVC Tubing Vacuum Valve A b Sample Holder Vacuum Valve B 2D Mechanical Stage Nozzle Vacuum Cube Optical Table Sample Holder Figure 2.3 Schematic view of vacuum system. Top panel: Front view of manifold. Bottom panel: Side view. (figure from [25]) 9

22 molecules, so that one does not have to worry about long time changes in the composition of fragments. (b) The average free propagation length of each molecule is much longer than the diameter of the molecular beam. This is especially important as it means that fragments and formed radicals due to laser excitation do not have time to react with other molecules, such that detected ions have to be derived from only one parent molecule. The laser beam (or in pump-probe experiments, both beams) enters the chamber through a thin (250 µm) sapphire window, so that the additional group velocity dispersion is negligible. Before that the beams are focused by a lens. It is adjusted such that the focal point is located within the molecular beam. 2.3 The red-red pump-probe alignment The pump-probe experiments use both arms of the Mach-Zehnder interferometer. Pulses from both arms are focussed in an non-collinear setup inside the chamber. This means that inside the crossing volume the phase difference between both pulses is not well defined. In particular this explains the absence of observable interference fringes. For each two pulse experiment, the system has to be aligned, such that spatial overlap of both pulses inside the chamber is established. For this pur- 10

23 pose a lens after the chamber maps the laser focal point inside the chamber to a white paper card. Here the optimal position for spatial overlap and maximal detector yield is marked, so that its is easy to reproduce the alignment. In order to adjust temporal overlap of both pulses, a similar setup is used. The imaging lens after the chamber images the focal point inside the chamber into a frequency mixing crystal (BBO: β BaB 2 O 4 ) cut for second harmonic generation. This crystal emits blue light between both frequency doubled red beams, if and only if both pulses arrive in the chamber at the same time. The intensity of the unshaped pulse is fixed for each run, but can be varied manually by opening and closing one iris in the beam path. As the second pulse comes from the pulse shaper, there is the possibility to change its intensity automatically via computer control. After the unshaped pulse energy is adjusted manually with the irises and a common power meter, the computer can be set to vary the remaining parameters. The final result is a three dimensional data set, that saves the time-of-flight-mass-spectrum (4000 to 6000 sampling points) for each time difference between shaped and unshaped pulse (in a certain range and step size), and for each energy of the shaped pulse (from total to zero energy). 11

24 2.4 The detector and data acquisition Our time-of-flight-mass-spectrometer (TOFMS) is configured to detect only positive ions. The detector sits above the interaction point. The cations are accelerated by a an applied voltage difference (750 V) and can escape through a small hole into the detection tube. Finally they reach microchannel plates that detect the single particles. The signal output is connected to a digital oscilloscope integrated in one of our lab computers. The oscilloscope is triggered with the laser pulse and can sample the signal with a rate of up to 500MHz. A typical trace is shown in figure 2.4. The y-axis shows the detector signal (in an arbitrary unit) for each sampling point numbered from 0 (trigger time) till the end of the trace. So the sampling point corresponds to time difference between trigger time and detector signal and thus yields the flight time of a certain particle. Furthermore the accelerating field is roughly homogeneous and constant in time, so that this time is related to the particles charge-to-mass ratio q m. After a re-calibration of the x-axis from flight time to q, one can assign each peak to a m certain ion. As the types of possible ions is limited by the sample molecule (and perhaps contaminations from former experiments) the assignment is mostly unique for the experiments performed in this thesis, besides few ambiguities like X ++ 2 or X +. The length of the TOF tube is short enough, so that the 12

25 fragments of two succeeding pulses do not overlap. Special care has to be taken for the pressure inside the chamber. Although higher pressure would give a stronger signal, it has disadvantages: A high ion flux shortens the lift time of the detector, and to ensure that the electric field inside the accelerating tube is homogeneous - even with lots of ionized molecules inside it (space charge effects) - their total number should be small. By experience, it is known that the working pressure stated above avoids this problem. Furthermore one can remove water from the sample by cooling it below its melting point (but keep above the melting point of water) and then evacuate it, so that the number of uninteresting water ions is kept low. 2.5 Frequency Resolved Optical Gating Frequency Resolved Optical Gating (FROG) is a technique to characterize a given laser pulse and retrieve the electric field E(t). It is described in detail in [10, 13]. This is useful on a daily basis for checking the pulse duration out of the amplifier as well as for analyzing the pulses generated by the pulse shaper. We use a second harmonic generation FROG (SHG-FROG). It basically works as a Mach-Zehnder-Interferometer, with one variable arm length, and a second harmonic crystal at the intersection position. The second harmonic generated light (sum frequency generation of both bemas) is then analyzed by a spec- 13

26 Signal [arb.] time [5 ns] Figure 2.4 Example trace of the TOFMS of CH 2 I 2, taken by the detector and the oscilloscope. It shows the TOFMS signal in dependence on the sampling point of the oscilloscope. The oscilloscope took sampling points at a frequency of 200 MHz. As the electric field is homogeneous one can infer from the flight time to the charge-to-mass ratio of the particles. Finally after re-calibrating the x-axis from time to charge-to-mass ratio, the peaks can be assigned to certain fragments. 14

27 trometer. Finally the FROG yields the Signal I sig in dependence of ω and time delay τ (by varying arm length): I sig (ω, τ) = E(t)E(t τ) e ıωt dt 2 (2.1) It can be proved that with an iterative algorithm the original electric field E(t) is retrievable from I sig (ω, τ) (compare [38]). The absolute phase is not retrievable and the result is symmetric to time inversion. 2.6 The UV-Red setup Some experiments have been also performed with mixed UV/red pump-probe scans. The complete UV setup is shown in figure 2.5. The vacuum system itself (chamber, sample injection, and pumping) is constructed in the same way as the old setup described in chapter 2.1. Similarly the data acquisition software can be reused. So this chapter is constrained to describing the differences, i.e. UV femtopulse generation and solving several problems caused by different behavior of the optics for UV and red pulses. The UV pulses are generated via frequency tripling. This setup is drawn in cartoon 2.6. The pulses first pass a beta barium borate crystal (BBO 1) to double the light to blue (390 nm) by second harmonic generation. A red beam with a power of about 900 µj per pulse generates about 170 µj blue 15

28 Prism Compressor R P 0 S Figure 2.5 UV-Red setup: The red laser light is partly frequency tripled to UV by some nonlinear crystals and then split up via mirror S. The red pulses (780 nm) are compressed by removing group velocity dispersion (GVD) in the prism compressor. A reflection setup is used where each pulse passes the compressor twice and is finally picked up by a mirror P and then sent to the delay stage. The UV pulses (262 nm) are modulated by a pulse shaper. Both beams are finally recombined by the mirror R and focussed inside the chamber. The AOM in the pulse shaper, as well as the delay stage can be controlled by the lab software. 16

29 BBO 1 Calcite BBO 2 crystal 780 nm 390 nm 780 nm 263 nm 390 nm 780 nm τ ω splitting UV mirror Figure 2.6 Red to UV tripling setup. First the red pulses are frequency doubled to blue pulses in the BBO crystal 1. The calcite window compensates for the phase difference between both pulses. In crystal BBO 2 blue and red pulses are mixed to generate UV light. pulse light. The UV pulses are generated from the red and blue ones via sum frequency generation inside the second BBO crystal. Both beams are generated collinearly in the first crystal, but as they have different group velocities, the blue and red pulse do not overlap in time. Therefore a thin calcite crystal in between the two BBO crystals is used to compensate for that. So finally the beam consists of femtosecond pulses with central wavelengths of 780 nm, 390 nm and 263 nm. The maximum UV pulse intensity is about 8 µj. After the tripling part a fused silica mirror coated for UV reflection is used to split up both interesting components. As in the both beam lines only dielectric mirrors coated either for red or for UV reflection are used, the blue component quickly gets lost. The red pulse passes the splitting mirror (fused silica, reflection coated for UV) unaffected, which is checked with the FROG and an unshaped pulse before. There is a convenient extension package of LabVIEW designed for ultrafast laser physics modeling [34]. Calculations with these show that the 17

30 red beam will have accumulated some chirp (of about 1000 (fs) 2 ) before it arrives at the chamber. To compensate for this, the beam is sent through a prism compressor, like it is described in section 3.1. It was decided to use a prism compressor instead of a compressor grating, which would have higher losses. The prisms can be aligned roughly at the Brewster angle and so have smaller reflection losses. With formula 3.4, the accurate spacing between both prisms can be calculated. The Fresnel Formulae predict (which is affirmed by a control measurement) a loss of 13%, which is considered tolerable. Finally the reflection mirror behind the second prism is tilted slightly, so that the reflected beam is not completely collinear with the incident beam, and can be separated by a pick-off mirror. Following the compressor, several mirrors are set to send the beam towards the delay stage. The UV beam then enters the UV pulse shaper after the splitting mirror. The mirror is constructed similarly to the red pulse shaper, as described in chapter 2.1 (or in full detail in [28]). After the pulse shaper, the UV beam line is combined again with the red beam line. The same type of mirror as for separating both colors is used. A final lens with very short focal length (15 cm) focusses the beam inside the chamber. As the beam consists of two colors, the focal length for both components is different (about 2 cm) as well as their achromatic errors. To circumvent the latter problem a collinear setup is used and the beams are 18

31 sent through the center of the lens. To match the different focal lengths an additional slighting focussing lens system (about 2 m) is added to the red beam line. Since both the red and the UV pulse have to arrive at the focal point at roughly the same time, the stage is moved to the time-zero position before each experiment, at which both beam lines are accurately of the same length. For the adjustment on a daily basis, a similar procedure to the red setup is used (compare to section 2.3). The beam leaves the chamber through a second window and a focussing lens. This lens maps the focal point inside the chamber to a white card. Because of the different focal lengths for both colors, only the UV is in focus. Here the optimal position of UV spot is marked, for an optimal alignment. This helped to refind the optimal position. As both beams are collinear this also implies optimal alignment for the red beam and thus spatial overlap for both beams in the chamber. For finding temporal overlap (time-zero), a BBO as used for sum frequency generation (BBO 2 in figure 2.5) is inserted after the chamber. After that, a prism splits the beam into its color components. Normally only a red and a UV spot are visible. But if and only if both pulses arrive at the same time, the BBO is able to generate blue light and a blue spot in between appears (the sum frequency mixing red + blue = UV is the reverse effect to 19

32 the difference frequency mixing UV red = blue). So the standard procedure to find temporal overlap for both pulses is to move the stage as long as a blue spot flashes. 20

33 Chapter 3 Ultrafast Optics In the following chapter the theoretical background of some ultrafast optical effects is discussed, in sofar as it is needed for the described experiments and their interpretation. 3.1 Dispersion Control Since all ultrashort pulses are simultaneously broadband wave packets, the control over their dispersion while passing through different materials is crucial. The overall dispersion is most easily described in frequency space. Therefore the spectral intensity G(ω) and spectral phase Φ(ω) of a pulse are defined by 21

34 its electric field E(t): E(t) = e ıωt G(ω) dω (3.1) G(ω) = G(ω) e ıφ(ω) (3.2) To describe the overall spectral phase of the pulse, it is expanded as a Taylor series (around the carrier frequency ω 0 ): Φ(ω) =φ 0 +(ω ω 0 )φ 1 +(ω ω 0 ) 2 φ (3.3) with the expansion coefficients φ k (k =1,...); they are called the k th order dispersion coefficients. When the beam propagates through a material, the modification in its phase can easily be described by multiplication with a factor H(ω). Now the pulse has the overall spectral phase Φ(ω)H(ω). Because of the Fourier theorem this is equivalent to just adding the dispersion coefficients of the material to the spectral phase coefficients of the pulse. φ 0 and φ 1 are the absolute phase and the group velocity of the pulse. As in our lab the pulse duration is 30 fs or longer, and the total beam length inside of transparent materials is small, we can restrict ourselves to controlling the amount of second order dispersion (or group velocity dispersion: GVD). A pulse with positive (negative) second order phase is called a positively (negatively) chirped pulse. 22

35 Since most materials add positive chirp to a pulse, special devices, like prism compressors or gratings, are used to remove the dispersion. Figure 3.1 shows a schematic view of a prism compressor. The idea is to split up the beam into its different frequencies. Then the red light, which is ahead in time in a positively chirped pulse, is sent on a longer path than the blue components. The second prism is aligned antiparallel to the first one, so that it can recombine all components to one single beam again. As the red component has to cover a longer distance, it will lose some phase difference to the blue component. So this adds a negative chirp. A reflecting mirror directly after prism 2 can make the beam pass the setup a second time and so doubles the effect and reverses the spacing of the beam along the y-axis. By contrast to this negative chirp, the pulse has to travel through small pieces of glass inside the prisms, which adds some positive chirp depending on the material. In order to calculate the total amount of added chirp, both effects have to be taken into account. If the second prism is moveable, one can vary the path length of the pulse inside the prism glass and thus its corresponding positive chirp. So the total additional chirp (positive, inside the prism and negative between them) is tunable. Reference [3] lists the correct formula for the added second order 23

36 Figure 3.1 Schematic of a prism compressor made with a reflection mirror to double the beam path. Both prisms (ideally AR-coated) split and recombine the laser beam; as long as they are antiparallel they compensate each other exactly. The red light has a longer beam path than the blue, which corresponds to a negative chirp. But the short way inside the prisms (D 1 and D C ) adds a strong positive chirp to the pulse depending on their material. Its amount can easily be modified by moving the second prism in and out. (figure taken from [25]) dispersion. A sufficient approximative formula is: GVD [ ( ) ] 2 λ3 D d2 n dn 2πc 2 dλ 4L 2 1 (3.4) dλ λ=λ 0 where D =2(D 0 + D C ) is the total path length inside the prism (for two bounces) and L 1 is the spacing between them. It can be seen that the overall added GVD splits up into a positive chirp of the prism medium itself (first summand proportional to D) and the negative chirp for space between them (second summand, proportional to the spacing L 1 ). 24

37 3.2 Nonlinear optical effects Nonlinear effects play an important role in ultrafast laser physics. Particularly in our lab: Kerr-lens modelocking in the Ti:Sapphire oscillator, SHG in the FROG, alignment for pump-probe experiments and frequency tripling for the UV generation. In a medium the original electric field E of the incident beam is added to the electric field of the bound charges in the material. Therefore the electric displacement field D, together with electric polarization P is defined as: D = ɛ 0 E + P In the simplest approach, the Lorentz model, the response of the electrons is linear: P = ɛ 0 χe, with the material constant χ as the electric susceptibility. In this case D is linear in E. But with increasing electrical field strength, the anharmonicity of the potential of the bound electrons results in a non-linear relationship between P and E. In the general case, χ is dependent on E, which can be described by a Taylor expansion (see [6]): P = n P (n) = ɛ n χ (n) E n In general the expansion coefficient χ (n) is a tensor of rank n + 1 (although I have suppressed indices for clarity). Obviously the first order term leads to the well known linear optical behavior. 25

38 Second harmonic generation and sum frequency generation are fundamental for the work described in this thesis, and most important for the generation of UV light (see 3.2). These two effects are caused by a non-vanishing χ (2). For simplicity we restrict ourselves to two plain, linearly polarized, monochromatic waves with the frequencies ω 1 and ω 2, so that the total electric field is E = E 1 e ıω1t + E 2 e ıω2t +c.c.. Thus the above formula yields: P (2) =ɛ 0 χ (2) E 2 (3.5) =ɛ 0 χ [ (2) E1e 2 2ıω1t + E2e 2 2ıω2t +c.c. ] + ɛ 0 χ [ (2) 2E 1 E 2 e ı(ω 1+ω 2 )t +c.c. ] +ɛ 0 χ (2) [ 2E 1 E 2e ı(ω 1 ω 2 )t +c.c. ] +2ɛ 0 χ (2) [E 1 E 1 + E 2 E 2] (3.6) Obviously the electromagnetic field inside the medium now contains also components of the frequency 2ω 1 and 2ω 2 (second harmonic generation), as well as ω 2 + ω 1 (sum frequency generation), ω 2 ω 1 (difference frequency generation) and a constant field shift. The formula also yields that all these effects are proportional to the product of the involved electric fields. This means for the intensity of the second harmonic generated light I SHG I 1 I 1, and it follows for the intensity of the frequency added light : I sum I 1 I 2. However usually none or only one of these generated frequencies will be observable. The reason for this is the so called phase matching condition. For example the second harmonic wave is always in phase with its fun- 26

39 damental at the place of its generation. But, as both waves have different phase velocities inside the medium, the second harmonic will (in general) be out of phase with the second harmonic generated at a different spot. If a non-vanishing output of frequency added light is desired, the phase matching condition has to be satisfied (with ω 3 = ω 1 + ω 2 ): k 1 + k 2 = k 3 n 1 ω 1 + n 2 ω 2 = n 3 ω 3 (3.7) Figure 3.2 shows the quantum mechanical interpretation for the sum frequency generation. Two photons are absorbed to an intermediate level and one with ω 3 is emitted. From this point of view the frequency of the emitted photon is obviously the sum frequency in order to ensure energy conservation. In addition, momentum conservation directly yields the phase matching condition. For isotropic media with normal dispersion, eqn. 3.7 can never be true, since ω 3 > ω 2,ω 1 and n 3 > n 2,n 1. Theoretically phase matching can be achieved by using anomalous dispersion. But there is a much easier and more common way: Birefringent crystals naturally provide different indices of refraction for different polarizations. By carefully adjusting the crystal orientation, phase matching can be established. 27

40 ω 2 ω 1 ω3 Figure 3.2 Sum frequency generation: One atom absorbs two photons of frequencies ω 1 and ω 2 and then emits only one photon ω 3 to drop again to ground level. Energy conservation leads to ω 1 + ω 2 = ω 3 and momentum conservation to the phase matching condition: k 1 + k 2 = k Ultrafast coherent control and closed loop feedback with genetic algorithm The topic of ultrafast coherent control covers a very broad field in physics. It centers around the idea of controlling optical, physical and chemical processes using coherence effects in matter by ultrafast lasers. A general overview is given in [33, 39], and some implementations are described in [7, 8, 32]. Shaped laser pulses can be used to modify the transition probability between two certain levels of a molecule. Particulary, this thesis discussed, if and by how much the Cl 2 and I 2 yield from CCl 4 and CH 2 I 2 can be increased by 28

41 changing the pulse shape. Finding the ideal pulse shape is theoretically very difficult and easily surmounts available computing power. The experimental approach to this is to use a genetic algorithm (GA), as described in [2]. The basic idea is shown in figure 3.3. The GA is an iterative algorithm to find the optimal pulse shape starting with an unshaped pulse. The name is chosen in analogy to biological evolution. The goal is set in the GA software, to enhance the fitness of the signal. The fitness is a functional based on the detector signal, e.g. in the easiest case, just the total count of a certain ion. Similarly to evolution, each pulse shape is encoded in the software by its genes; normally 20 to 40 points, that give the desired phase of the laser pulse at certain frequencies. The complete pulse phase is given by an interpolation between these points. Each generation starts with the optimal pulse shapes of the last generation. From these parent pulses the child pulses are generated by two operations: crossover and random variation, which resemble the biological processes of recombination and mutation. The first one hybridizes the genes from a pair of parents to generate the child genes. The second operation varies the genes randomly around the starting genes. The child pulses are finally written via the pulse shaper to the laser pulse and tested in the interaction cube on the sample. The GA calculates the fitness for each pulse and then selects the best 29

42 pulse shapes to be the parents for the next generation. A graph showing the fitness versus iteration qualitatively looks like shown in figure 3.4. The fitness is staring to rise dramatically after a few iterations and finally saturates. It usually takes 50 to 100 iterations till the GA feedback converges to a pulse shape. As the GA uses random elements, two questions arise: As to(a) about the reproducibility of the results and (b) if the algorithm really finds global optimal solutions, instead of local maxima. To (a), principally there can be an infinite number of optimal results. But at least in the experiments described in this thesis, the GA results are grouped into several classes: the trivial result (enhancing fitness by removing higher order dispersion, which causes a stronger signal) and sets of mutually similar non-trivial solutions. As to (b), there exist theoretical articles (like [30]) that prove that GAs are principally capable to find optimal global solutions, although they assume ideal conditions, that are far from practical situations: that there are no limits for the GA in search space, i.e. no limits in pulse power, pulse duration, phase manipulation, etc. 30

43 Figure 3.3 Closed loop learning control: A set of pulse shapes is randomly generated and written to the laser pulse by the pulse shaper. These pulses are used for the experiments. The detector signal is then used by the GA to calculate the effectiveness of the pulse. The best pulses are then the starting point for the next loop. 31

44 Fitness [arb] Number of Iteration Figure 3.4 Qualitative evolution of the fitness (averaged over the top 8 pulse shapes) as a function of generation. (Here optimized for Cl + 2 out of CCl + 4 ) 32

45 Chapter 4 CCl 4 experiments The first molecule studied was Carbon-tetrachloride (CCl 4 ). We examined the laser driven reaction: CCl 4 +nhν Cl (4.1) This was done for various motivations: First a former group member, D. Cardoza, detected Cl + 2 while working with CCl 4. Also there are some papers (like [23] and [5]) observing this process under femtolaser excitation and to form Cl 2 (neutral or ionic) from CCl 4. The chemical structure of the molecule is shown in figure 4.1. The goal of these experiments is to verify the results of other groups, to check if this process also produces chlorine ions, and if so, what the involved molecular levels are and if the Cl + 2 yield can be controlled. For this purpose shaped single pulse experiments are performed in combination with a genetic 33

46 Figure 4.1 Schematic view of a CCl 4 molecule. It has a methane-like shape, whereby the hydrogen atoms are replaced with chlorine atoms. algorithm. 4.1 General facts of the Cl 2 data The sample had several non-ideal properties. It was naturally contaminated with Cl 2 from the very beginning on, both because of the way it is industrially produced (MP Biomediacls gives a purity of 99.9%), and because of the sensitivity of halogenated methanes to air and light. Once opened, the chemical aged and formed molecular chlorine inside the storage bin. Furthermore, there was contamination from former experiments in the chamber. Thus the measured yield of Cl + 2 was only partly due to Cl + 2 formed from CCl 4 by interaction with the laser. In order to determine what fraction of the mea- 34

47 sured Cl + 2 was due to contamination we made measurement to estimate the Cl 2 contamination. We found that approximately half of the detected Cl + 2 is due to background. This make its an open question, whether the observed pulse shape depending behavior is due to the Cl + 2 formed from CCl 4 or due to the background Cl 2. There were two possible ways for estimating the number of Cl + 2 in the chamber. The first was to use the TOFMS signal and the second one was to use a residual gas analyzer (RGA). When the laser intensity was high enough, the signal of all fragments started to saturate (as a function of pulse energy), which was due to the fact that finally all molecules in the laser beam were ionized and fragmented in some way. This happened at about 200 µj, as can be seen in figure 4.2. In this case, each fragmented CCl 4 molecule produced exactly one ion that contained a C atom. The signal sum of CCl CCl C + was proportional to the overall number of fragmented CCl 4 molecules inside the beam. So the sum signal of these fragments at a sufficiently high energy (when all these particles are already in the saturating regime) gave the number of CCl 4 molecules inside the laser focus (times a constant, that represents the detection efficiency). The exact value of this constant is not needed, as the detector sensitivity for each ion was equal and one only looked at relative signal. 35

48 Ion Yield [arb.] Cl CCl 3 + CCl 2 CCl + C + Energy Scan for CCl Energy [µj] Figure 4.2 Ion yield (in arbitrary units) of fragmented CCl 4 for pulse energies from 0 to 210µJ: obviously CCl + 3 has the biggest contribution. At the maximum energy 210µJ all fragments have started to saturate. Their values are: 8400 (CCl + 3 ), 1400 (CCl + 2 ), 2300 (CCl + ), 200 (both for C + and Cl + 2 ) 36

49 Similarly the Cl + 2 signal was a lower bound for the Cl 2 inside the beam, either formed by fragmentation or just ionized contamination. It was just a lower bound, as some Cl 2 was likely to be destroyed by the laser. For the example in figure 4.2 the situation for the highest pulse energy 210 µj can be read. As CCl 4 has no stable parent ion, the CCl + 4 signal is zero. So this leads to the following estimate: N(formedCl 2 andcontamination) N(parentmolecules) = > 1.62% (4.2) A different approach is to use the RGA of one of the chambers. Since the parent ion is not stable, fragmented ions were observed just by ionizing CCl 4. The electron impact mass spectrum [21] affirmed that just ionizing it does not produce any Cl 2. So in this case Cl + 2 was neither formed from the parent molecule nor destroyed by the laser, and the RGA just showed the partial pressure of the background chlorine. And again summing up over all fragments that contain a carbon gave the partial pressure of CCl 4. So by assuming that the sensitiveness for all particles is roughly equal, this leads to an estimate for the contamination: N(Cl 2 contamination) N(parentmolecules) 0.8% (4.3) This was below the lower bound (4.2). This clearly proves the formation of Cl + 2 from CCl + 4 by the laser. 37

50 But one has to consider that in the worst case half of the detected chlorine signal was due to contamination. In other words the signal from contamination Cl 2 was comparable to the signal from generated chlorine. Therefore we could not rule out that all GA results are not just background effects, although most likely effects from both contamination and newly formed Cl 2 were observed. The most direct solution to this problem would have been to repeat all these experiment with pure Cl + 2. Then we could have distinguished the signal from contamination or created Cl + 2. But for several reasons, we did not perform these experiments. First a practical problem: Our vacuum system is built up to insert liquid samples and to work with its low vapor pressure. We would have had to construct a new vacuum setup for it and do this very carefully for security hazards. And second, there were worries about our system itself, as parts of the vacuum system, especially the turbo pump and seals, could be corroded. This has been the main reason for continuing with studying CH 2 I 2, as I 2 is much easier to handle. 4.2 Genetic Algorithm Experiments These experiments studied whether the Cl + 2 yield can be increased by changing the shape of the pulse interacting with the molecule. Therefore a single red pulse was used and optimized by the GA. Figure 4.3 shows the TOFMS of 38

51 CCl 4 produced by an unshaped pulse. We fed back on the Cl + 2 signal for a variety of pulse energies and GA parameters. In the naive approach the fitness functional of the algorithm was the total Cl + 2 yield. But also the control over a normalized Cl + 2 yield was examined, i.e. the quotient or difference to other fragments (with appropriate weight factors). As the parent ion is not stable, CCl + 2 and CCl + 3 were alternatives. The diverse runs found optimal solutions, that increased the Cl + 2 yield up to 8.0% at maximal pulse energy (around 200µJ) and up to 100% for low energies (around 45µJ). Figure 4.4 compares the TOFMS signal of a shaped and an optimized pulse, taken at a pulse energy of 40 µj. The most successful pulses were scanned by the FROG and their shape was reconstructed. An example is shown in shown in figure 4.5. A common feature of all optimal pulse shapes, even for different fitness functionals, was a three intensity peak structure, with linear phases during the peaks. The time distances between the peaks varied, but were always between 50 fs and 150 fs. Furthermore, we examined the ratio of optimized Cl + 2 yield to unshaped pulse Cl + 2 yields in dependence of the pulse energy. This revealed the interesting behavior shown in figure 4.6. The effectiveness of the GA control seems to be energy dependent and shows a hyperbolic-like behavior. The relative difference in efficiency was highest for low pulse energies and decreased for 39

52 Figure 4.3 TOFMS of CCl4at 220µJ pulse energy: The relevant peaks are labeled with the corresponding fragments. There is no CCl + 4 peak, as the parent ion itself is not stable. Because of different Cl isotopes (35 amu and 37 amu), several ions have multiple peaks. Fragment Mass [amu] CCl2 140 C Cl2 150 CCl C Cl3 Signal [arb.] 180 N2 Cl H2O

53 [arb.] Figure 4.4 TOFMS of CCl 4 with an optimized pulse compared to an unshaped laser pulse at an energy of 40 µj. The box zooms in to the interval of the Cl + 2 signal. Here the Cl + 2 signal is enhanced by about 70%. higher energies. From a very pessimistic point of view this behavior could be just the GA increasing the pulse peak intensity (by removing higher order dispersion). This would be a trivial behavior and not of interest. But one can show that this effect is at least not the dominating effect: Otherwise, if there had been a significant amount of higher order dispersion, which the GA just had compensated for, this would have caused a constant factor for the ratio shaped/unshaped peak energy. According to graphic 4.7 some of the data was taken in a range of energy µj, in which the Cl + 2 yield shows a linear dependence on the pulse energy. Thus, at least in this range, compensation of higher order dispersion would have contributed only a constant factor 41

54 Figure 4.5 Example pulse shape with maximal Cl + 2 yield: The pulse shape was measured with the FROG. The graph shows the temporal intensity (blue) and temporal phase (green). All optimal pulses had a three peak structure in common. 42

55 for all energies. As the graph 4.6 is obviously not constant in this interval ( µj), dispersion compensation could not be the main effect to cause this behavior. The energy dependent efficiency of the GA indicated the presence of different mechanisms for producing Cl + 2. With low energy, only few of these channels were possible and control over these (or only one of them) seemed to be easy and effective. With higher energies additional mechanisms became available, which responded in a different way to the shaped pulses. So finally the averaged effect on all channels together by the shaped pulse was less impressive. As described in section 4.1 it remained unclear if these different channels only comprise Cl + 2 formation from CCl 4 or also include Cl + 2 production from ionizing background Cl 2. 43

56 Figure 4.6 Relative Cl + 2 yield of a GA improved shaped pulse to an unshaped pulse at different energies. The fitness functional was set to maximize the Cl + 2 ion yield. 44

57 relative Signal Cl2+ CCl3+ CCl Energy [ µj] Figure 4.7 Energy scan of CCl 4 : TOFMS signal of different fragments at different energies of the unshaped pulse: Cl + 2 (black), CCl + 3 (red), CCl + 2 (green). Each graph is normalized to 1 at its maximum. There is an almost linear dependence on energy in the Cl + 2 signal between 70 µj and 100 µj 45

58 Chapter 5 Diiodomethane CH 2 I 2 experiments The CH 2 I 2 experiments are the main experiments of my thesis. There are several papers that claim a similar behavior for CH 2 I 2 to form I 2 under femtolaser excitation, like [23, 24, 43]. The article [44] especially deals with the possible underlying mechanisms for the concerted elimination of neutral I 2, and [26] shows that the efficiency of this process depends on the chirp of the pulse. [22] is of special interest as it explicitly addresses the contamination problem that has already occurred with CCl + 4 and studies this reaction with a laser systems comparable to ours. In addition, CH 2 I 2 has several advantages to CCl 4. The generally weaker bonds of iodine to carbon than chlorine to carbon promise a more effective breaking process. Figure 5.1 shows the arrangement of the atoms in this molecule. 46

59 Figure 5.1 Structure of the CH 2 I 2 molecule: The tetrahedron-like structure of methane is stretched due to the heavy (black) iodine atoms. The (grey) carbon atom is centered and bound to the (white) hydrogen atoms. 47

60 Generally CH 2 I 2 samples suffer also from the same contamination problem as CCl 4, as they naturally contain molecular iodine. Similar to CCl 4 the sample ages when exposed to air and light, which increases the I 2 contamination. The first try with a commercially available sample affirmed this apprehension, as it contained about 1% molecular iodine. Luckily Professor Marecek from the Chemical Synthesis Center at Stony Brook University supplied us with a mostly iodine free sample. All experiments described in this chapter use this sample that is additionally stabilized with a copper wire, which binds molecular iodine chemically inside the storage bin. A further advantage of this molecule is that iodine is much less aggressive than chlorine and solid at room temperature, which make it possible to perform control experiments with pure iodine, as I will present later. So finally in the next sections it will be demonstrated that it is possible to distinguish the iodine signal caused by contamination from the signal caused by iodine derived from CH 2 I 2 by this way. Three types of experiments were performed with this molecule: (a) genetic algorithm experiments (similar to these with CCl 4 ) to examine, if it is possible to control the I + 2 yield, (b) pump probe experiments with two red laser pulses and (c) pump probe experiments with a UV and a red pulse. 48

61 5.1 Genetic Algorithm experiments This section will be discuss if the iodine yield from fragmentation of CH 2 I 2 can be enhanced by different pulse shapes. These experiments are the counterpart to the GA experiments with CCl 4 described in chapter 4. The ion yields of different fragments versus pulse energy (unshaped, transform limited pulse) are plotted in figure 5.2. With these numbers it is possible to roughly estimate the expected background iodine signal, similarly to the section about the chlorine contamination 4.1. If the laser pulse energy is high enough all molecules inside the laser field are ionized or even fragmented, which can be seen in the saturation of the signal with high pulse energy. In particular each excited CH 2 I 2 molecule generates exactly one ion, that contains one C atom. So the total number of CH 2 I 2 (times a constant corresponding to the detector sensitivity) in the interaction area can be estimated by adding all ion signals: C +,CH +,CH + 2,..., CH 2 I + 2. Reference numbers can be found in figure 5.2 at 210 µj maximum energy, at which these fragments have reached the saturation yield. With the numbers given in this plot (the ion yield for CH and C is not shown but negligible), one can guess the total ion signal due to CH 2 I 2 in the chamber is The first commercial sample contains 1% iodine, as stated by its producer Acros Organics, which leads to an expected ion signal of 70 or lower from 49

62 8000 Ion yield [arb.] I+ I2+ CH2I2+ CH2I+ N+ and CH Energy [µj] Figure 5.2 ion yield (in an arbitrary unit) of fragmented CH 2 I 2 for (unshaped) pulse energies from 0 to 210µJ: At the maximum energy 210µJ all fragments have started to saturate. Their values are: 2170 (CH 2 I + 2 ), 3660 (CH 2 I + ), 330 (I + 2 ), 1500 (N + and CH + 2 ). iodine contamination. Figure 5.2 also shows that the I + 2 ion signal at maximum pulse energy is about 330, which is much more than the expected signal of 70 due to iodine contamination. This proves that at least 79% of the detected iodine is due to iodine formation from CH 2 I 2. Background iodine ions contribute less than 21% to the signal. As the second sample, provided by Professor Marecek, is much cleaner, one could expect this contribution to be even lower. The GA experiments themselves are performed in a similar way as the GA experiments with CCl 4. The algorithm is set to search for optimal solution for different fitness functions, like maximum I + 2, or maximum ratios and differences 50

63 1 5 Intensity [arb] Phase [rad] Time Delay [ps] Figure 5.3 Optimal pulse shape for maximum I + 2 with a CH 2 I 2 sample. The shape of one main peak and several smaller ones following is characteristic. to I + 2 and other fragments (CH 2 I + 2,CH 2 I +,I + ). A typical GA enhanced pulse shaped is shown in figure 5.3 and can increase the iodine yield up to 12% compared to an unshaped pulse. All optimal pulses showed a similar behavior: One big intensity peak followed by a few smaller ones. The time difference between the first and second peak is about 55 fs, and between the following peaks about 40 fs. The phase is constant during an intensity peak and has jumps in between. The significant amount of signal (21%) due to background I 2 raises again the question, whether the enhancement is due to more effective I + 2 splitting from CH 2 I + 2 or due to more effective ionization of background iodine. Therefore 51

64 the same GA runs are repeated with a pure iodine sample. Indeed, the GA is able to improve the I + 2 yield from a pure iodine sample. But its improvement does not surpass 8.3%. Together with the information that the iodine contamination is at most 21%, it is fair to say that the iodine contamination contributes at most 1.7% to the total I + 2 GA enhancement in the CH 2 I 2 sample. This is an upper bound that is only reached, if the optimal pulse for I + 2 formation from CH 2 I 2 is at the same time an optimal pulse for background iodine ionization. This leaves an increase of at least 10.3% for I + 2 formation from CH 2 I 2. So finally it can be concluded from this section that (a) I + 2 can be formed from CH 2 I 2 by laser excitation and (b) the efficiency of this process can be enhanced by at least 10.3% by modifying the pulse shape. It remains an open question whether this increase is significant and it is based on molecular dynamics rather than dispersion compensation. Eventually the pump-probe experiments described in the next chapter will provide more information on the internal molecular dynamics. 5.2 Red-Red pump probe experiments The red-red pump probe experiments provide the main data of this thesis and the basis of the interpretation of the mechanism to form I + 2 in chapter 6. The 52

65 preparations for the setup are described in section 2.1. These measurements are made both with a CH 2 I 2 sample as well as a pure I 2 sample in order to differentiate between laser driven concerted elimination and simply ionization of residual I 2 in our CH 2 I 2 sample. The most interesting piece of data is presented in figures 5.4 and 5.5. Both are color coded plots of ion yield for different fragments as a function of pump/probe delay and pulse energy. Note that the pulse, whose arriving time is varied, is the pulse with constant energy. This energy is kept constant at 110 µj for both scans shown. The arrival time of the pulse whose energy is varied is kept constant. The time difference between both pulses is varied from 4000 fs to 4000 fs in 9.9 fs steps, which means that the pulses switch their role during the scan. The second pulse energy is varied from 0 µj to 100 µj, with a resolution of 21 steps. So on the left hand side the sample is pumped first with the strong 110 µj pulse and then probed with the energy varied second pulse. On the right hand side, the sample is pumped with the energy varied pulse first and then probed with the 110 µj pulse. As both scans are taken at a long time scale and the vapor pressure of the pure iodine has not reached a equilibrium, these plots are already pressure corrected. Therefore the linear slope in the ion yields is subtracted. Characteristically, all pump-probe scans show a strong time-zero peak. 53

66 This is due to overlap of pump and probe pulses and the nonlinear response of the molecules to the laser intensity. One could expect interference patterns, but both laser pulses have no fixed phase difference throughout the interaction region. 5.3 UV-Red pump probe experiments Similar data was taken with a UV-red mixed pump-probe setup, as described in chapter 2.6. The result of the most interesting scan is plotted (pressure corrected) in figure 5.6. Here the pump probe delay is varied again within the range from 4000 fs to 4000 fs in 9.9 fs steps. The red unshaped pulse energy is kept constant during the whole scan at 55 µj. The UV pulse energy is varied automatically from 0 µj to1.4 µj in 100 steps. Similar to the red pump-probe experiments, the original data shows a linear decrease in pressure. But due to leaks in the vacuum system between the sample holder and the reservoir at 3 times the pressure suddenly increases (once on the left side, twice on the right side). As it is known from former experiments in the group the water signal shows no time dependent dynamics, it is used as a indicator for the current pressure. So in this case the data is pressure corrected by normalizing it with a spline interpolation of the water 54

67 (a) H 2 O + signal (b) I + 2 signal (c) CH 2 I + 2 signal (d) CH 2I + signal Figure 5.4 These figures show the ion yields for different fragments in a redred pump-probe scan with a CH 2 I 2 sample. The 2D-plot is color coded in dependency on the shaped pulse energy and time delay. On the left hand side the energy-constant 110 µj pulse comes first, on the right hand side the energy-varied pulse comes first. 55

68 (a) I + 2 signal (b) H 2O + signal Figure 5.5 These figures show the ion yields for different fragments in a redred pump-probe scan with a pure iodine sample. The 2D-plot is color coded in dependency on the shaped pulse energy and time delay. On the left hand side the energy-constant 110 µj pulse comes first, on the right hand side the energy-varied pulse comes first. signal. 56

69 (a) H 2 O + signal (b) I + 2 signal (c) CH 2 I + 2 signal (d) CH 2I + signal Figure 5.6 These figures show the ion yields for different fragments in a UV-red pump-probe scan with a CH 2 I 2 sample. The red pulse energy is constant at 55 µj), the UV pulse energy is varied from 0 µj to1.4 µj. On the left hand side the red pulse comes first, on the right hand side the UV pulse comes first. 57

70 Chapter 6 Discussion of the CH 2 I 2 pump-probe data In this chapter the pump-probe experiments, described in sections 5.2 and 5.3 are interpreted. With it the mechanism driving the I + 2 emission from CH 2 I + 2 can be quantitatively resolved. 6.1 The oscillations in the ion yield All pump probe scans (e.g. in the figures 5.4, 5.5 and 5.6) obviously show oscillations in the CH 2 I + 2,CH 2 I + and I + 2 signal. First of all, one can rule out technical causes (e.g. power fluctuations in the laser, detector signal processing), as this would affect the signal of all fragments. It is known from earlier experiments that the water signal does not to show any molecular dynamics, except the time zero peak. As it can be seen in 5.4(a), 5.5(b) and 5.6(a) the water signal behaves as predicted and is constant in time, except the usual 58

71 noise. Thus the oscillations have to be caused by a time changing interaction behavior of the fragments themselves with the laser field. The molecule cannot be in an eigenstate of the field free Hamiltonian when it interacts with the second laser pulse. The eigenstates of each molecule can be characterized by electronic states, vibrational quantum numbers and rotational quantum numbers. [36] states time periods longer than 1.5 ps for rotational motion and ( ) fs for vibrational excitation in CH 3 I, which are roughly comparable to CH 2 I 2.AsCH 2 I 2 has a larger angular momentum, one can expect the rotational motions to be even slower. So the timescale for rotational coherence is much longer than the observed oscillations. Finally electronic transitions happen on a sub-femtosecond timescale and thus are much to fast to be observed. This indicates that the observed behavior is due to a vibrational wave packet. Therefore these oscillations will be the key point to resolve the molecular dynamics. 6.2 The left hand side of the red-red data The two sides of the pump probe scan shown in figures 5.4 and 5.5 demonstrate principally different behavior: On the left hand side the pump energy is kept constant, whereas the probe 59

72 ion yield [arb.] CH 2 I + 2 CH 2 I + I + 2 mean (red 110 µj - red 55 µj) LHS oscillations (red 110 µj - red 55 µj) mean (red 55 µj - red 110 µj) RHS oscillations (red 55 µj - red 110 µj) Table 6.1 Ion yield and oscillation amplitude (in per cent of the CH 2 I + yield on the LHS) of some fragments (averaged in time) for two red pulses red (55 µj and 110 µj) in different time order. energy is varied. So the group of states accessed by the pump is constant, which leads to qualitatively constant dynamics in energy. But on the right-hand side the pump energy is varied, the probe energy is kept constant. In this case different states are excited with varying the energy, which finally leads to different dynamics for different energies. This fact makes the right hand side more difficult to interpret. Additionally, as it can be seen from table 6.1, the strongest oscillations are observed at the left hand side, with a first pulse (at fixed 110 µj) and a weaker second pulse afterwards. As this part of the data will be the most conclusive in interpreting the I + 2 formation process and understanding the observed oscillations, the discussion will be focussed on it. 60

73 In order to take a closer look at the observed oscillations, some fragment signals are Fourier transformed along the time axis. They are plotted versus frequency (in wave numbers) and probe energy in panel 6.1. The frequency resolution is ±4cm 1. This figure clearly demonstrates that the CH 2 I + 2,CH 2 I + and I + 2 derived from CH 2 I 2 share oscillations at 111 cm 1. The I + 2 signal in the CH 2 I 2 sample shows two Fourier components: one at 111 cm 1, which is absent for probe pulse energies below 8 µj and a second at 130 cm 1, which is dominant at low probe energies. The I + 2 signal from a pure iodine sample oscillates with 130 cm 1 at low probe energy. This contribution continually decreases in favor of a component oscillating at 121 cm 1 with increasing probe energy. The fact that the Fourier spectra do not change in energy beyond 50 µj (see in fig. 6.1), neither in phase nor in amplitude (as it can be seen when the line outs are plotted separately), allows to average over the signals for this energy range. The result can be found in figure 6.2, where the ion yield versus pump-probe delay is drawn. The graphs are normalized in order to compare the yields for different fragments. 61

74 (a) (b) (c) (d) Figure 6.1 Panels: (a) I + 2 (from CH 2 I 2 ), (b) CH 2 I + 2, (c) CH 2 I +,I + 2 (from I 2 ). They show the Fourier transformation (in time) of these fragment signals in a pump-probe scan, versus frequency and probe energy. The pump energy was constant at 110 µj, the probe energy was varied from 0 µj to90µj in 21 steps. 62